Nvidia races ahead in AI hardware with new Rubin chips

Nvidia has made another major move in the global race toward high-end AI solutions. The company says its new Rubin data center chips are nearly ready for launch later this year, and customers will soon begin testing them. 

All six Rubin chips have returned from Nvidia’s manufacturing partners, Nvidia said. These chips have already undergone several key testing procedures and are on schedule to be used by customers. The firm notes that this early progress is important because AI companies and data centers are racing to get faster and more efficient hardware. 

Nvidia Chief Executive Officer Jensen Huang shared the update in a keynote address at the Consumer Electronics Show (CES) in Las Vegas. CES is among the largest technology gatherings in the world, at which manufacturers unveil new products and ideas. “The race is on for AI,” Huang said amid the rapidly developing race among technology companies to unlock the next generation of AI. 

His comments indicate Nvidia is very sure it can still be the world’s number one manufacturer of AI accelerators. These accelerators are specialized chips that a data center uses to train and run AI models, which power tools such as chatbots, image recognition systems, and advanced robots.

Rubin chips deliver a major performance boost

Rubin is Nvidia’s newest AI accelerator, following its previous generation called Blackwell. The company says Rubin is 3.5 times faster at training AI models and five times faster at running AI software compared to Blackwell. Training AI involves teaching models to learn from vast amounts of data, while running AI means utilizing those models to perform tasks in real-time.

Rubin adds a new central processing unit (CPU) with 88 cores. Cores are the parts of a chip that perform calculations and process data. With twice the performance of the chip it replaces, this new CPU is better suited for more complex AI workloads. At Nvidia’s spring GTC conference in California, the company typically shares full product details. 

This time, more information was sold out than usual. The move is considered a means of keeping consumers and developers focused on Nvidia’s hardware as AI adoption continues to grow rapidly. Huang himself has also made numerous public appearances promoting AI products, partnerships, and investments. Nvidia wasn’t the only player in the spotlight at CES. Lisa Su, CEO of rival chipmaker Advanced Micro Devices (AMD), also booked a keynote, highlighting escalating competition in the chip market.

Nvidia courts big customers as competition rises

Some investors have worried that competition for Nvidia is heating up. Other tech companies are also developing their own AI chips, making it difficult to say whether spending on AI can keep pace. 

Nvidia, however, has been upbeat, believing the long-term AI market could be worth trillions of dollars, driven by demand from industries such as cloud computing, businesses, and emerging sectors.

The Rubin hardware will be used in Nvidia’s DGX SuperPod, a powerful supercomputer designed for large-scale AI work. At the same time, customers will be able to buy the Rubin chips as individual components, allowing them to build more flexible and modular systems.

The increased performance is particularly critical, given that AI systems continue to evolve. Modern AI increasingly relies on networks of specialized models that not only process massive amounts of data but also solve problems in multiple steps. Such tasks include planning, reasoning, and decision-making. 

Nvidia also emphasized that Rubin-based systems will be cheaper to operate than Blackwell systems. Because Rubin can deliver the same results with fewer components, data centers can save on energy and operating costs. 

Major cloud computing companies such as Microsoft, Google Cloud, and Amazon Web Services (AWS) are expected to be among the first to deploy Rubin hardware in the second half of the year. These companies currently account for the majority of spending on Nvidia-powered AI systems.

Sharpen your strategy with mentorship + daily ideas - 30 days free access to our trading program